Abrishambaf, Amin

Ultra-high-performance concrete (UHPC) possesses mechanical characteristics that significantly outperform traditional concrete. However, replicating these properties consistently across different production batches—even when using the same recipe—remains a challenge.

This dataset examines how variations in raw materials, environmental conditions, dosage variation, and both mixing and curing practices affect the mechanical properties of UHPC produced from a single reference formulation. Designed according to a three-phase design of experiments methodology, the dataset comprises 150 systematically planned experiments, offering a comprehensive view of the multiple factors influencing UHPC quality. Measurements of compressive and flexural strengths are provided at 24 hours and after 28 days post-mixing. Beside the mechanical properties, the dataset includes five characteristics of the fresh state, measured directly after each mixing process.

All experiments are conducted in the laboratory of G.tec Engineering GmbH under controlled conditions, using the same mixer, same mixing tool, and the same team of technicians. The environment is maintained at a constant temperature of 20 °C throughout the experimental process.

From each experiment, three specimens are cured under the designed conditions. First, the flexural strength is measured by carefully halving each of the three specimens. Then, the resulting six halves are used to measure the compressive strength. Finally, after a careful analysis of the results from each specimen, the averages for flexural and compressive strengths are reported.

The dataset also includes outliers. After analysis by UHPC experts and the data science team, 11 data points (numbered 5, 17, 30, 36, 41, 47, 57, 99, 101, 128, and 148) were identified and removed as outliers to assure data quality.

By offering a structured collection of high-dimensional data and a relatively small data size, this dataset is particularly suitable for advanced regression analyses, notably those addressing sparse data scenarios.